## What color is your Jacobian? Graph coloring for computing derivatives (2005)

Venue: | SIAM REV |

Citations: | 44 - 10 self |

### BibTeX

@ARTICLE{Gebremedhin05whatcolor,

author = {Assefaw Hadish Gebremedhin and Fredrik Manne and Alex Pothen},

title = {What color is your Jacobian? Graph coloring for computing derivatives},

journal = {SIAM REV},

year = {2005},

volume = {47},

pages = {629--705}

}

### Years of Citing Articles

### OpenURL

### Abstract

Graph coloring has been employed since the 1980s to efficiently compute sparse Jacobian and Hessian matrices using either finite differences or automatic differentiation. Several coloring problems occur in this context, depending on whether the matrix is a Jacobian or a Hessian, and on the specifics of the computational techniques employed. We consider eight variant vertexcoloring problems here. This article begins with a gentle introduction to the problem of computing a sparse Jacobian, followed by an overview of the historical development of the research area. Then we present a unifying framework for the graph models of the variant matrixestimation problems. The framework is based upon the viewpoint that a partition of a matrixinto structurally orthogonal groups of columns corresponds to distance-2 coloring an appropriate graph representation. The unified framework helps integrate earlier work and leads to fresh insights; enables the design of more efficient algorithms for many problems; leads to new algorithms for others; and eases the task of building graph models for new problems. We report computational results on two of the coloring problems to support our claims. Most of the methods for these problems treat a column or a row of a matrixas an atomic entity, and partition the columns or rows (or both). A brief review of methods that do not fit these criteria is provided. We also discuss results in discrete mathematics and theoretical computer science that intersect with the topics considered here.

### Citations

4331 |
Computer Architecture: a quantitative approach (3rd ed
- Hennessy, Patterson
- 2003
(Show Context)
Citation Context ...uld be considerably larger in size. A larger graph size in turn implies that more storage space is required and performance is degraded due to the hierarchical structure of memory in modern computers =-=[60]-=-. In section 3.6, we provide computational evidence to support this claim. A third advantage of the bipartite graph–based formulation is related to flexibility. Unlike the column intersection graph, t... |

2196 | Numerical Optimization - Nocedal, Wright - 2006 |

952 |
Approximation Algorithms
- Vazirani
- 2001
(Show Context)
Citation Context ...gister allocation in compilers [24], radio and wireless networks [44, 84], scientific computing [51, 73, 74, 111], data movement in distributed and parallel computing [58], facility location problems =-=[117]-=-, cache-efficient algorithms [77], etc. Parallel computers make it feasible to solve large-scale problems in many of these application areas, especially optimization, and hence there is curGRAPH COLO... |

551 | Matching Theory - Lovasz, Plummer - 1986 |

381 | A simple parallel algorithm for the maximal independent set problem - Luby - 1986 |

373 |
New methods to color the vertices of a graph
- Brélaz
- 1979
(Show Context)
Citation Context ...(SD) ordering, have also been suggested as heuristics for reducing the number of colors used by a greedy algorithm. The former ordering is due to Coleman and Moré [30] and the latter is due to Brélaz =-=[22]-=-. ID and SD orderings are similar in spirit to LF ordering; they differ in the specialized “degree” measure they use. Unlike LF, SL, and ID orderings, an SD ordering can be computed only as the colori... |

363 |
2000), Evaluating derivatives: principles and techniques of algorithmic differentiation, number 19
- Griewank
(Show Context)
Citation Context ...ingle color distinct from the colors assigned to other nonzeros, and then all nonzeros of the same color are grouped into one compressed column. (Figure 1.2 is inspired by similar figures in Griewank =-=[53]-=- and Hovland [67].) 0 0 0 0 0 0 0 0 0 0 0 j 55 c 5 c 5 c 4 c 4 c 1 c 1 c 3 c 3 c 2 c 2 r 1 r 2 r 3 r 4 r 5 r 1 r 2 r 3 r 4 r 5 c 1 c 2 c 3 c 4 c 5 c 1 c 2 c 3 c 4 c 5s634 A. H. GEBREMEDHIN,F. MANNE,AN... |

304 |
Graph Coloring Problems
- Jensen, Toft
- 1995
(Show Context)
Citation Context ...stance-1 coloring, a discussion of the coloring number of a graph, and a brief review of theoretical results on some of the coloring problems addressed in this paper. The monograph by Jensen and Toft =-=[71]-=- provides a detailed account of over 200 open coloring problems and an extensive list of references. The survey article by Toft [115] has a good discussion of mathematical results for the distance-1 c... |

295 |
Every Planar map is Four Colorable
- Appel, Haken, et al.
- 1977
(Show Context)
Citation Context ...st known as the four color theorem, is perhaps one of the most famous results in graph theory. After challenging mathematicians for nearly a century, the theorem was proved by Appel and Haken in 1976 =-=[7, 10]-=-. However, due to its length, its extensive use of computer verification, and its omission of details, the proof of Appel and Haken has been subject to criticism. In response, the authors published a ... |

222 |
Exploratory Social Network Analysis with Pajek
- Nooy, A, et al.
- 2005
(Show Context)
Citation Context ...ite [88]. (A graph is k-degenerate if every induced subgraph has minimum degree at most k.) A related concept has been studied in the social and biological network literature as the k-core of a graph =-=[21, 105]-=-. The k-core of a graph is a maximal induced subgraph in which every vertex has at least k neighbors in the subgraph. A k-core of a graph can be computed by a linear-time algorithm that visits the ver... |

203 | Free Bits, PCPs and Non-Approximability – Towards Tight Results
- Bellare, Goldreich, et al.
- 1995
(Show Context)
Citation Context ...pproximation ratio for the distance-1 coloring problem log log |V | is O(|V |( log |V | )3 ) [59]. Also, the problem is known to be not approximable within O(|V | 1/7−ɛ ) for any ɛ > 0, unless P = NP =-=[12]-=-. Despite these rather pessimistic results, there exist several practically effective distance-1 coloring heuristics [30, 94]. A detailed discussion of some of these heuristics is included in section ... |

203 |
Register allocation via coloring
- Chaitin, Auslander, et al.
- 1981
(Show Context)
Citation Context ...ing derivative matrices with substitution methods. Graph and hypergraph coloring have been used in a wide collection of application areas in addition to optimization: register allocation in compilers =-=[24]-=-, radio and wireless networks [44, 84], scientific computing [51, 73, 74, 111], data movement in distributed and parallel computing [58], facility location problems [117], cache-efficient algorithms [... |

190 |
Computing the minimum fill-in is NP-complete
- Yannakakis
- 1981
(Show Context)
Citation Context ...our or more vertices has a chord, an edge connecting two nonconsecutive vertices on the cycle. Computing a chordal completion of a graph in which the number of fill edges |F | is minimized is NP-hard =-=[124]-=-. This graph problem is a well-known model for sparse Cholesky factorization. The process through which G ′ is computed resembles, but is not the same as, the elimination game, a process through which... |

189 | Recipes for Adjoint Code Construction
- Giering, Kaminski
- 1998
(Show Context)
Citation Context ...e reverse mode requires a row-partitioning into the fewest possible groups of structurally orthogonal rows. A number of AD software tools are available, including ADIFOR [16], Odyssee [110], and TAMC =-=[50]-=- for FORTRAN 77, and ADIC [17] and ADOL-C [55] for C/C++. Some AD tools, such as ADIFOR and ADIC, support a vector mode in which several bidirectional derivatives are computed simultaneously. Using su... |

182 | Approximate graph coloring by semidefinite programming
- Karger, Motwani, et al.
- 1998
(Show Context)
Citation Context ...o O(n( (log log n)2 Berger and Rompel [14]. The current best known approximation ratio is O(n (log n) 3 ), a result due to Halldórsson [59]. Based on Wigderson’s algorithm, Karger, Motwani, and Sudan =-=[75]-=- considered approximate graph coloring using semidefinite programming techniques. On the negative side, Bellare, Goldreich, and Sudan [12] showed that distance-1 coloring is not approximable within O(... |

168 | ADOL-C, a package for the automatic differentiation of algorithms written in C/C
- Griewank, Juedes, et al.
- 1996
(Show Context)
Citation Context ...o the fewest possible groups of structurally orthogonal rows. A number of AD software tools are available, including ADIFOR [16], Odyssee [110], and TAMC [50] for FORTRAN 77, and ADIC [17] and ADOL-C =-=[55]-=- for C/C++. Some AD tools, such as ADIFOR and ADIC, support a vector mode in which several bidirectional derivatives are computed simultaneously. Using such tools, the compressed derivative matrix (se... |

122 |
Estimation of sparse Jacobian matrices and graph coloring problems
- Coleman, Mor·e
- 1983
(Show Context)
Citation Context ...o estimate the Jacobian. They described a simple greedy algorithm for partitioning the columns of a Jacobian into a small number of groups of structurally orthogonal columns. In 1983 Coleman and Moré =-=[30]-=- modeled this matrix partitioning problem as a distance-1 graph coloring problem. The model uses the column intersection graph of a matrix where columns correspond to vertices and two vertices are joi... |

122 | Efcient routing in all-optical networks
- Raghavan, Upfal
- 1994
(Show Context)
Citation Context ...t of colors to paths in a graph serves as a theoretical model for akk aℓℓ ⎟ ⎠ .sGRAPH COLORING FOR COMPUTING DERIVATIVES 661 routing problems. Examples of works that discuss such applications include =-=[1, 108]-=-, and a recent survey is available in [15]. The term star coloring owes its name to the fact that in such a coloring, every subgraph induced by vertices assigned any two colors is a collection of star... |

107 |
On Acyclic colorings of planar graphs
- Borodin
- 1979
(Show Context)
Citation Context ...3. Grünbaum [56] also proved that every planar graph admits an acyclic coloring using at most nine colors. After a series of improvements [4, 81, 98] the number was finally reduced to five by Borodin =-=[19]-=-. Borodin’s proof is similar to the proof of the four color theorem in that there are 450 reducible configurations that need to be checked. Star coloring has been recently studied by Albertson et al. ... |

107 |
Jacobian-free Newton-Krylov Methods: A survey of approaches and applications
- Knoll, Keyes
(Show Context)
Citation Context ...ing purposes, and only a subset of the matrix elements needs to be computed. Computing a good preconditioner is critical for fast convergence to a solution. A recent survey article by Knoll and Keyes =-=[79]-=- discusses various applications where the “Jacobian-free Newton–Krylov” method is used. A basic ingredient of this method is an approximate computation of some elements of the Jacobian. Also, there ar... |

99 |
Osprey: a network visualization system
- Breitkreutz, Stark, et al.
- 2003
(Show Context)
Citation Context ...ite [88]. (A graph is k-degenerate if every induced subgraph has minimum degree at most k.) A related concept has been studied in the social and biological network literature as the k-core of a graph =-=[21, 105]-=-. The k-core of a graph is a maximal induced subgraph in which every vertex has at least k neighbors in the subgraph. A k-core of a graph can be computed by a linear-time algorithm that visits the ver... |

97 |
Some lower bounds for a class of frequency assignment problems
- Gamst
- 1986
(Show Context)
Citation Context ...itution methods. Graph and hypergraph coloring have been used in a wide collection of application areas in addition to optimization: register allocation in compilers [24], radio and wireless networks =-=[44, 84]-=-, scientific computing [51, 73, 74, 111], data movement in distributed and parallel computing [58], facility location problems [117], cache-efficient algorithms [77], etc. Parallel computers make it f... |

96 |
An upper bound to the chromatic number of a graph and its application to time-table problems
- Welsh, Powell
- 1967
(Show Context)
Citation Context ...ng of the vertices with respect to their degrees in the input graph G minimizes the upper bound in Theorem 11.3. This ordering, known as largest first (LF) ordering, was suggested by Welsh and Powell =-=[120]-=-. In general, however, the number of colors used by SEQ using an LF ordering can be much larger than the optimal number of colors. An example of a bipartite (hence 2-colorable) graph with n vertices a... |

89 |
Every planar graph is 5-choosable
- Thomassen
- 1994
(Show Context)
Citation Context ...ndently by Chiba, Nishizeki, and Saito [25]. Also, a linear-time algorithm that uses five colors is implicit in a recent proof of the theorem “every planar graph is five colorable” given by Thomassen =-=[114]-=-. Deciding whether a planar graph of maximum degree four can be distance-1 colored using three colors is NP-complete [71]. Wegner [119] studied distance-k (k ≥ 2) chromatic numbers of planar graphs ha... |

87 |
ADIFOR 2.0: Automatic differentiation of Fortran 77 programs
- Bischof, Carle, et al.
- 1996
(Show Context)
Citation Context ...an efficient application of the reverse mode requires a row-partitioning into the fewest possible groups of structurally orthogonal rows. A number of AD software tools are available, including ADIFOR =-=[16]-=-, Odyssee [110], and TAMC [50] for FORTRAN 77, and ADIC [17] and ADOL-C [55] for C/C++. Some AD tools, such as ADIFOR and ADIC, support a vector mode in which several bidirectional derivatives are com... |

86 |
Acyclic Colorings of planar graphs
- Grunbaum
- 1973
(Show Context)
Citation Context ...lic coloring) called triangular coloring. The acyclic coloring formulation is due to Coleman and Cai [27], who called the coloring variant cyclic. The name acyclic coloring was introduced by Grünbaum =-=[56]-=-, the first to define and study the problem within an entirely different context. A variant of a substitution method for computing a Jacobian matrix has been considered by Hossain and Steihaug [63]. T... |

84 | ADIC — An Extensible Automatic Differentiation Tool for ANSI-C
- Bischof, Roh, et al.
- 1996
(Show Context)
Citation Context ...partitioning into the fewest possible groups of structurally orthogonal rows. A number of AD software tools are available, including ADIFOR [16], Odyssee [110], and TAMC [50] for FORTRAN 77, and ADIC =-=[17]-=- and ADOL-C [55] for C/C++. Some AD tools, such as ADIFOR and ADIC, support a vector mode in which several bidirectional derivatives are computed simultaneously. Using such tools, the compressed deriv... |

84 |
On the estimation of sparse Jacobian matrices
- Curtis, Powell, et al.
- 1974
(Show Context)
Citation Context ...r 4 r 5 r 1 r 2 r 3 r 4 r 5 c 1 c 2 c 3 c 4 c 5 c 1 c 2 c 3 c 4 c 5s634 A. H. GEBREMEDHIN,F. MANNE,AND A. POTHEN Fig. 1.2 A Jacobian matrix and its compressed representation. Curtis, Powell, and Reid =-=[35]-=- were the first to observe in 1974 that sparsity can be employed in this way to reduce the number of function evaluations needed to estimate the Jacobian. They described a simple greedy algorithm for ... |

79 | Efficient routing and scheduling algorithms for optical networks
- Aggarwal, Bar-Noy, et al.
- 1994
(Show Context)
Citation Context ...t of colors to paths in a graph serves as a theoretical model for akk aℓℓ ⎟ ⎠ .sGRAPH COLORING FOR COMPUTING DERIVATIVES 661 routing problems. Examples of works that discuss such applications include =-=[1, 108]-=-, and a recent survey is available in [15]. The term star coloring owes its name to the fact that in such a coloring, every subgraph induced by vertices assigned any two colors is a collection of star... |

79 |
Recursive blocked algorithms and hybrid data structures for dense matrix library software
- Elmroth, Gustavson, et al.
(Show Context)
Citation Context ...he fact that a column intersection graph–based approach requires larger memory size, which in turn implies larger memory access time due to the multilevel memory hierarchies in modern microprocessors =-=[39]-=-. A similar explanation can be offered for the observation that building a bipartite graph takes more time than distance-2 coloring it afterwards (see columns TGb and Tcol of Table 3.2). It should be ... |

78 | A Still Better Performance Guarantee for Approximate Graph Coloring
- Halldorsson
- 1993
(Show Context)
Citation Context ... a polynomial time algorithm, the algorithm is usually referred to as a heuristic. The current best known approximation ratio for the distance-1 coloring problem log log |V | is O(|V |( log |V | )3 ) =-=[59]-=-. Also, the problem is known to be not approximable within O(|V | 1/7−ɛ ) for any ɛ > 0, unless P = NP [12]. Despite these rather pessimistic results, there exist several practically effective distanc... |

70 | A parallel graph coloring heuristic
- Jones, Plassman
- 1993
(Show Context)
Citation Context ...rgraph coloring have been used in a wide collection of application areas in addition to optimization: register allocation in compilers [24], radio and wireless networks [44, 84], scientific computing =-=[51, 73, 74, 111]-=-, data movement in distributed and parallel computing [58], facility location problems [117], cache-efficient algorithms [77], etc. Parallel computers make it feasible to solve large-scale problems in... |

70 |
Optimal approximation of sparse hessians and its equivalence to a graph coloring problem
- McCormick
- 1983
(Show Context)
Citation Context ...rtis, Powell, and Reid to compute sparse Hessians. Their approach exploits the fact that only one of each pair of symmetric off-diagonal elements in the Hessian matrix needs to be computed. McCormick =-=[97]-=- introduced a distance-2 graph coloring model for the computation of Hessians in 1983. The model uses the adjacency graph representation of the underlying symmetric matrix and requires that in every p... |

69 | On chromatic number of graphs and set systems
- Erdős, Hajnal
- 1966
(Show Context)
Citation Context ... color G. Motivated by this fact, the quantity B ∗ (G) + 1 is known in the literature as the coloring number of G and is denoted by col(G). The name coloring number was first used by Erdös and Hajnal =-=[40]-=-. As we will shortly see, the graph parameter B ∗ (G) is also of interest in contexts other than coloring. The coloring number may appear to be an intractable graph parameter as the minimum is evaluat... |

67 |
Improving the performance guarantee for approximate graph coloring
- Wigderson
- 1983
(Show Context)
Citation Context ... on n vertices, the first approximation algorithm for distance-1 coloring, with an approximation ratio of O(n 1 ), was obtained by Johnson [72]. This log n log log n log n )2 log log n ) by Wigderson =-=[121]-=- and to O(n( log n )3 )by was later improved to O(n( (log log n)2 Berger and Rompel [14]. The current best known approximation ratio is O(n (log n) 3 ), a result due to Halldórsson [59]. Based on Wigd... |

57 | ILUM: A multi-elimination ILU preconditioner for general sparse matrices
- Saad
- 1996
(Show Context)
Citation Context ...rgraph coloring have been used in a wide collection of application areas in addition to optimization: register allocation in compilers [24], radio and wireless networks [44, 84], scientific computing =-=[51, 73, 74, 111]-=-, data movement in distributed and parallel computing [58], facility location problems [117], cache-efficient algorithms [77], etc. Parallel computers make it feasible to solve large-scale problems in... |

49 | Coloring Powers of planar graphs
- Agnarsson, Halldórsson
- 2000
(Show Context)
Citation Context ... (k ≥ 2) chromatic numbers of planar graphs having maximum degree at most three and proved that for k = 2 eight colors suffice, whereas seven is the best possible. Recently, Agnarsson and Halldórsson =-=[3]-=- studied distance-k coloring of planar graphs. They showed that a planar graph having a sufficiently large maximum degree ∆ can be distance-2 colored using at most ⌈9∆/5⌉ colors. The authors also show... |

44 | On Algorithms for Efficient Data Migration
- Hall, Hartline, et al.
- 2001
(Show Context)
Citation Context ...s in addition to optimization: register allocation in compilers [24], radio and wireless networks [44, 84], scientific computing [51, 73, 74, 111], data movement in distributed and parallel computing =-=[58]-=-, facility location problems [117], cache-efficient algorithms [77], etc. Parallel computers make it feasible to solve large-scale problems in many of these application areas, especially optimization,... |

43 |
Estimation of sparse Hessian matrices and graph coloring problems
- Coleman, Moré
- 1984
(Show Context)
Citation Context ...ing finite differences, in addition to exploiting sparsity using structurally orthogonal partitions, the number of function evaluations can be reduced further by exploiting symmetry. Coleman and Moré =-=[31]-=- later defined the following more general version of the partition underlying the approach of Powell and Toint. Definition 4.2. A partition of the columns of a symmetric matrix A is symmetrically orth... |

42 | Software for estimating sparse Jacobian matrices - COLEMAN, GARBOW, et al. - 1984 |

40 |
Acyclic colorings of graphs, Random Structures and Algorithms 2
- Alon, McDiarmid, et al.
- 1991
(Show Context)
Citation Context ...tisfies the inequality χa(G) ≤ ∆ + 1. A negative answer to this question was given by Erdös, who proved probabilistically the existence of graphs where χa(G) ≥ ∆ 4/3−ɛ [71]. Alon, McDiarmid, and Reed =-=[6]-=- showed (again by probabilistic methods) that there exists a constant c1 > 0 such that χa(G) ≤ c1 ·∆ 4/3 . The same authors also showed that there is a constant c2 > 0 such that there exist graphs wit... |

38 | Algorithms for square roots of graphs
- Lin, Skiena
- 1995
(Show Context)
Citation Context ...efficient programs, including parallel ones, for the distance-2 coloring problem need to be developed. 3.5. Algorithms. For every fixed integer k ≥ 1, the distance-k graph coloring problem is NP-hard =-=[89]-=-. Thus, in practice, one is bound to rely on approximation algorithms or heuristics. An algorithm A is said to be a γ-approximation algorithm fors650 A. H. GEBREMEDHIN,F. MANNE,AND A. POTHEN Algorithm... |

38 |
Estimation of sparse Jacobian matrices
- Newsam, Ramsdell
- 1983
(Show Context)
Citation Context ...s produced from a partition of a set of columns or rows (or both) of a matrix; i.e., a set is divided into groups in which an element of the set belongs to one and only one group. Newsam and Ramsdell =-=[103]-=- suggested a method for computing a Jacobian that relies on solving an overdetermined rectangular system of equations, a system produced from a column grouping that is not necessarily a partition. The... |

36 | Scalable iterative solution of sparse linear systems
- Jones, Plassmann
- 1994
(Show Context)
Citation Context ...rgraph coloring have been used in a wide collection of application areas in addition to optimization: register allocation in compilers [24], radio and wireless networks [44, 84], scientific computing =-=[51, 73, 74, 111]-=-, data movement in distributed and parallel computing [58], facility location problems [117], cache-efficient algorithms [77], etc. Parallel computers make it feasible to solve large-scale problems in... |

35 | The Efficient Computation of Sparse Jacobian Matricies Using Automatic Differentiation, Cornel Theory Center
- Coleman, Varma
- 1996
(Show Context)
Citation Context ...g problem. Their formulation relies on the notion of a consistent row-column partition in which the entire set of rows and columns is partitioned into two respective sets of groups. Coleman and Verma =-=[32]-=- also studied the same problem and identified a similar bidirectional partitioning problem. Their notion of partition differs from that of Hossain and Steihaug in that it partitions only a subset of t... |

35 |
Issacson, “Graph coloring algorithms
- Matula, Marble, et al.
- 1972
(Show Context)
Citation Context ...nown to be not approximable within O(|V | 1/7−ɛ ) for any ɛ > 0, unless P = NP [12]. Despite these rather pessimistic results, there exist several practically effective distance-1 coloring heuristics =-=[30, 94]-=-. A detailed discussion of some of these heuristics is included in section 11 of this paper. In the current section we show how a distance-1 coloring heuristic can be easily adapted to the distance-2 ... |

34 |
Graph Theory, 2nd ed
- Diestel
- 2000
(Show Context)
Citation Context ...e k-core to a hypergraph and discusses the k-cores of protein interaction graphs and protein complex hypergraphs. The coloring number of a graph is also closely related to the arboricity of the graph =-=[37]-=-.s696 A. H. GEBREMEDHIN,F. MANNE,AND A. POTHEN 11.3. Bounds for the Chromatic Number. Since computing the chromatic number χ(G) of a general graph G is NP-hard, we consider bounding it with more tract... |

34 |
On the estimation of sparse Hessian matrices
- Powell, Toint
- 1979
(Show Context)
Citation Context ...sification of partitioning/coloring problems arising in sparse derivative matrix computation. Problems P1 through P8 are addressed in this paper. 1.4.1. Jacobian vs. Hessian. In 1979 Powell and Toint =-=[107]-=- extended the approach of Curtis, Powell, and Reid to compute sparse Hessians. Their approach exploits the fact that only one of each pair of symmetric off-diagonal elements in the Hessian matrix need... |

34 |
An inequality for the chromatic number of a graph
- Szekeres, HS
- 1968
(Show Context)
Citation Context ...+ 1. The quantity on the right side of the inequality in the preceding statement is precisely BSL(G) + 1. Thus we have col(G) ≥ BSL(G) +1, which completes the proof. Matula [93] and Szekeres and Wilf =-=[113]-=-, independently, have related the coloring number of a graph to yet another seemingly intractable graph parameter. The parameter we are referring to is the maximum minimum degree in an induced subgrap... |

33 |
The cyclic coloring problem and estimation of sparse Hessian matrices
- Coleman, Cai
- 1986
(Show Context)
Citation Context ...nd Moré [31]. The latter authors in particular found a less accurate model (in comparison with acyclic coloring) called triangular coloring. The acyclic coloring formulation is due to Coleman and Cai =-=[27]-=-, who called the coloring variant cyclic. The name acyclic coloring was introduced by Grünbaum [56], the first to define and study the problem within an entirely different context. A variant of a subs... |